Re: [tip:core/rcu] rcu: Remove inline from forward-referenced functions

["Paul E. McKenney" <paulmck@linux.vnet.ibm.com>]

On Tue, Nov 10, 2009 at 05:03:41PM -0800, Joe Perches wrote:
> On Tue, 2009-11-10 at 15:41 -0800, Paul E. McKenney wrote:
> > On Tue, Nov 10, 2009 at 03:06:20PM -0800, Joe Perches wrote:
> > > On Tue, 2009-11-10 at 14:59 -0800, Paul E. McKenney wrote:
> > > > > non-inline functions in .h files probably aren't a good idea.
> > > > ;-)
> > > > Here are my options:
> > > []
> > > > Please note that kernel/rcutree_plugin.h is internal to RCU -- only
> > > > kernel/rcutree.c includes it, so there is no possibility of conflicting
> > > > definitions.
> > > > Any options that I am missing?
> > > Maybe something like:
> > > #ifdef whatever
> > > #define RCU_ANNOUNCE "Hierarchical RCU implementation.\n"
> > > ...
> > > #else
> > > #define RCU_ANNOUNCE "Experimental preemptable hierarchical RCU implementation.\n"
> > > ...
> > > #endif
> > > Use pr_info(RCU_ANNOUNCE) instead of rcu_bootup_announce();
> > > ?
> > This would still be a forward reference, right?
> 
> No, what I suggest would delete the forward reference and
> the  rcu_bootup_announce function altogether.

I still either have a plugin definition in the non-plugin file
kernel/rcutree.c or a forward reference to the macro definition.

> The single caller of rcu_bootup_announce in __rcu_init
> would become printk(KERN_INFO RCU_ANNOUNCE);
> 
> I wrote the original quoted content above without looking
> at the file, just the submitted patch.  After looking the
> the rcutree files, another option is to do what Ingo did
> with sched.c and finesse the #include ".h".
> 
> $ grep "include.*\.c" sched.c
> #include "sched_idletask.c"
> #include "sched_fair.c"
> #include "sched_rt.c"
> # include "sched_debug.c"
> 
> So maybe rename rcutree_plugin.h to rcutree_plugin.c and
> #include "rcutree_plugin.c" in rcutree.c instead.

Hmmm...

							Thanx, Paul